Meta has introduced a new approach to training its artificial intelligence systems by leveraging internal employee activity data, including mouse movements and keystrokes. The initiative is designed to provide AI models with real-world examples of how users interact with computers in everyday tasks.

According to the company, the goal is to improve the functionality of AI-powered assistants by teaching them to better replicate human behavior in digital environments. By analyzing patterns such as clicking, navigation, and menu selection, Meta aims to enhance the responsiveness and efficiency of its AI tools.

A company spokesperson explained that the data collection will be limited to specific applications and will include safeguards to protect sensitive information. Meta also stated that the data gathered will be used բացառively for training purposes and not for any unrelated activities.

The development highlights the growing demand for high-quality training data within the AI industry, where companies are increasingly exploring unconventional sources to refine their models. Access to realistic user interaction data is considered critical for building systems capable of performing complex digital tasks autonomously.

However, the move raises fresh concerns about privacy and data usage, particularly as companies expand the scope of information collected for AI development. Recent reports have indicated that some firms are also mining historical corporate communications to further enhance their models.

As the competition in artificial intelligence intensifies, Meta’s strategy underscores the lengths technology companies are willing to go to secure valuable training data, while also prompting ongoing debate about ethical boundaries in data collection.

Source: TechCrunch